Palm Oil and Rubber Price and Trader’s Behavior at International towards Local Level

(1)

Palm Oil and Rubber Price and Trader’s Behavior at International towards Local Level

Dissertation to obtain the Ph.D. degree

in the International Ph.D. Program for Agricultural Sciences in Goettingen (IPAG) at the Faculty of Agricultural Sciences,

Georg-August-University Goettingen, Germany

presented by Rakhma Melati Sujarwo born in Bontang, Indonesia

Goettingen, May 2020

(2)

(3)

D7

Name of supervisor: Prof. Dr. Bernhard Brümmer

Name of co-supervisor: Prof. Dr. Stephan von Cramon-Taubadel Name of co-supervisor: Prof. Dr. Oliver Mußhoff

Date of Oral Examination: May 20, 2020

(4)

(5)

i

Acknowledgements

Praise to Allah the Almighty for all His blessings. Having the opportunity to study in Georg-August University Göttingen, Germany gave me one of the most precious experiences in my life. I feel grateful to finally finish my dissertation with all the supports I received from all parties. First and foremost, I would like to express my deepest appreciation to my supervisor Prof. Dr. Bernhard Brümmer. Without his guidance this dissertation would not have been possible. His trust, patience and support are priceless so that I can complete this study with all the shortcomings. Having the opportunity to be part of his team at the Chair of Agricultural Market Analysis has contributed greatly to my career development.

Also, I acknowledge the support and help of Prof. Dr. Stephan von Cramon-Taubadel and Prof. Dr. Oliver Mußhoff. in co-supervising this dissertation and being part of the examination committee.

Thanks to my office mate, Yashree, who is an expert in so many things in life. What am I without you? Also, thanks to all other chair members (Tim, Gabriel, Claudia, Tom, Yueming, Yuan, Enrique, Oliver, Bernhard, Jurij, Dela, Nina, and Nina Enke) who give joy and happiness on the 10^th floor of the Blue Tower. And to all parties as part of the Collaborative Research Centre 990, Efforts Project. This project, under DFG, has also provided financial support during my study. Special thanks go to all counterparts in IPB University (Dr. Dedi Budiman Hakim and the late Prof. Dr. Rina Oktaviani) and Jambi University (Prof. Dr. Zulkifli Alamsyah and Dr. Mirawati Yanita), as well as my team who helped me during my data collection in Jambi.

Furthermore, my deepest gratitude and super big hug go to my lovely family who always support and encourage me with their dedication during my study: Ayah and Ibu who have always supported the path of life that I have chosen. Also, my lifetime partner, the best dad in the world for our children, who has never left me in supporting my study activities.

Salma, the first, who always bears witness to my sadness and happiness in undergoing this process. And Musa, the second, who appeared to bring joy at the critical moments in the completion of this dissertation.

Last, but not least, I am indebted to my many friends and colleagues who supported and gave me invaluable relationships, from PPI Göttingen, Kalam Göttingen, and all my new international friends in Germany.

--- and this Corona virus pandemic situation shall never be forgotten ---

(6)

ii

List of Tables

Table 2. 1 The definitive EU AD rate for Indonesian biodiesel producers. ... 14

Table 2. 2 Summary Statistics of Price Variables (USD/kg) ... 23

Table 2. 3. Log-likelihood Values of Restricted and Unrestricted Models ... 26

Table 3. 1 Number of All Active Traders in the Survey Area ... 35

Table 3. 2 Transition Matrix of Traders in Period 1 (2012-2015) ... 35

Table 3. 3 Transition Matrix of Traders in Period 2 (2015-2018) ... 35

Table 3. 4 Average Income of Traders Existing both in 2012 and 2015 (million IDR) ... 36

Table 3. 5 Average Income of Traders Existing both in 2015 and 2018 (million IDR) ... 36

Table 3. 6 Definition of Variables ... 37

Table 3. 7 Number of All Active Traders in the Survey Area ... 41

Table 3. 8 Result of Logit and Marginal Effect Estimation ... 44

Table 3. 9 New and Existing Traders’ Characteristics in 2018 ... 48

Table 4. 1 Definition of Variables ... 60

Table 4. 2 Attributes, Levels and Descriptions ... 64

Table 4. 3 An example of a Choice Set (simplification) ... 64

Table 4. 4 Village and Larger Traders’ Characteristics ... 65

Table 4. 5 Credit Providers and Traders not Providing Credit Characteristics ... 66

Table 4. 6 Estimation Statistics of All Model Estimations ... 68

Table 4. 7 Log-likelihood Ratio Test ... 69

Table 4. 8 Parameter Estimates from the ML_2 Model Estimation ... 70 Table 4. 9 Rubber Traders’ Willingness to Charge Price Reduction for Rubber Attributes71

(9)

v

List of Figures

Figure 2. 1 Net CO2 Life Cycle Emissions of Petroleum Diesel and Biodiesel Blends ... 6

Figure 2. 2 Biodiesel and Renewable Diesel (HVO) in the European Union ... 7

Figure 2. 3 AD Initiations between 1995-2018. ... 10

Figure 2. 4 Dumping Illustration ... 11

Figure 2. 5 Value of Export Tariff (USD/ton) of Indonesian CPO and Biodiesel ... 12

Figure 2. 6 Indonesian Biodiesel Dumping Illustration ... 13

Figure 2. 7 Home Equilibrium without Trade, with Trade and with AD in Trade ... 15

Figure 2. 8 The AD duty Illustration Effect in Indonesia ... 16

Figure 2. 9 Share of Indonesian CPO Export Value based on Importing Countries ... 21

Figure 2. 10 Share of Indonesian CPO Export Value based on Ports’ Location ... 22

Figure 2. 11 Area and Production of Oil Palm Plantation in the Jambi Province ... 22

Figure 2. 12 Time Series Plot of Price Variables ... 23

Figure 2. 13 Value of Export Tariff and Levy (USD/ton) of Indonesian CPO ... 24

Figure 2. 14 Time Series Plot with Structural Break ... 25

Figure 2. 15 Export Value of Biodiesel from Indonesia to the EU (USD thousand) ... 27

Figure 2. 16 Visualization of VECM estimation between 𝑃𝐼𝐷 and 𝑃𝑊 ... 28

Figure 2. 17 Visualization of VECM estimation between 𝑃𝐼𝐷 and 𝑃𝑊 ... 29

Figure 3. 1 Price Volatility of Palm Oil and Rubber Price (2008-2019) ... 32

Figure 3. 2 Map of Jambi ... 41

Figure 3. 3 Illustration of the Increase of Probability per Unit Variable... 45

Figure 3. 4 Illustration of the Increase of Probability per Unit Interacted Variable... 47

Figure 3. 5 Frequency of Existing and New Traders by Traded Product ... 48

Figure 4. 1 Number of Farmers by Ownership ... 51

Figure 4. 2 Probability Design in Unrelated Response Randomized Response Technique 55 Figure 4. 3 Illustration of rubber trader individual preferences ... 57

(10)

vi

List of Appendices

Appendix 2. 1 Unit Root Tests ... 84 Appendix 2. 2 Johansen Cointegration Test between Indonesian and World CPO Prices . 84 Appendix 2. 3 Johansen Cointegration Test between Jambi FFB and Indonesian CPO Prices ... 84 Appendix 2. 4 Gregory-Hansen Cointegration Test between Indonesian and World CPO Prices ... 85 Appendix 2. 5 Gregory-Hansen Cointegration Test between Jambi FFB and Indonesian CPO Prices ... 85 Appendix 2. 6 Johansen Cointegration Test between Indonesian and World CPO Prices before the Breakpoint ... 86 Appendix 2. 7 Johansen Cointegration Test between Indonesian and World CPO Prices after the Breakpoint ... 86 Appendix 2. 8 Johansen Cointegration Test between Jambi FFB and Indonesian CPO Prices before the Breakpoint ... 86 Appendix 2. 9 Johansen Cointegration Test between Jambi FFB and Indonesian CPO Prices after the Breakpoint ... 86 Appendix 2. 10. VECM Estimation between Indonesian and World CPO Prices ... 87 Appendix 2. 11 VECM Estimation between Indonesian and World CPO Prices allowing Structural Break ... 88 Appendix 2. 12 VECM Estimation between Indonesian and World CPO Prices before the Breakpoint ... 89 Appendix 2. 13 VECM Estimation between Indonesian and World CPO Prices after the Breakpoint ... 90 Appendix 2. 14 VECM Estimation between Jambi FFB and Indonesian CPO prices ... 91 Appendix 2. 15 VECM Estimation between Jambi FFB and Indonesian CPO allowing Structural Break ... 92 Appendix 2. 16 VECM Estimation between Jambi FFB and Indonesian CPO before the Breakpoint ... 93 Appendix 2. 17 VECM Estimation between Jambi FFB and Indonesian CPO after the Breakpoint ... 94

Appendix 3. 1 Name of Regencies, Districts and Villages of Survey Location ... 95

(11)

vii

Appendix 3. 2 Stata Output of Variable Descriptive Statistics Year 2012... 96

Appendix 3. 5 Stata Output of Logit Estimation Year 2012 ... 98

Appendix 3. 6 Stata Output of Marginal Effect Estimation Year 2012 ... 99

Appendix 3. 7 Stata Output of Logit Estimation Year 2015 ... 100

Appendix 3. 8 Stata Output of Marginal Effect Estimation Year 2015 ... 101

Appendix 3. 9 Stata Output of Specification Error Year 2012 ... 102

Appendix 3. 10 Stata Output of Specification Error Year 2015 ... 102

Appendix 3. 11 Stata Output of Collinearity Diagnostics Year 2012 ... 103

Appendix 3. 12 Stata Output of Collinearity Diagnostics Year 2015 ... 104

Appendix 3. 13 Stata Output of Wilcoxon Rank-sum (Mann-Whitney Test of Farming Revenue by Traders’ Status ... 105

Appendix 3. 14 Stata Output of Trading Revenue Summary Statistics by Remain Year 2015 ... 106

Appendix 3. 15 Stata Output of Trading Revenue Summary Statistics Year 2012 and 2015 ... 106

Appendix 3. 16 Wilcoxon rank-sum (Mann-Whitney) non-parametric test ... 107

Appendix 3. 17 Pearson Chi2 test ... 108

Appendix 4. 1 A Choice Set ... 109

Appendix 4. 2 Descriptive Statistics of Respondents (n=210) ... 110

Appendix 4. 3 Stata Output of Shapiro-Wilk Normality Test ... 111

Appendix 4. 4 Stata Output of Wilcoxon Rank-sum (Mann-Whitney) Test by Traders’ Status (left) and Credit Provision Status (right) ... 112

Appendix 4. 5 Stata Output of Spearman's Rank Correlation Coefficients ... 113

Appendix 4. 6 Results across Different Model Estimations ... 114

Appendix 4. 7 Stata Output of CL_1 Model Estimation ... 115

Appendix 4. 8 Stata Output of CL_2 Model Estimation ... 116

Appendix 4. 9 Stata Output of ML_1 Model Estimation ... 117

Appendix 4. 12 The Cholesky Factorization of the Covariance Matrix ... 120

Appendix 4. 13 Stata Output of the ML_3 Random Coefficients Covariance Matrix ... 120

(12)

viii Appendix 4. 14 Summary of Rubber Traders’ WTC Price Reduction for All Attributes

across Different Model Estimations ... 121

Appendix 4. 15 WTC Stata Output of CL_1 Model Estimation ... 122

Appendix 4. 16 WTC Stata Output of CL_2 Model Estimation ... 122

Appendix 4. 17 WTC Stata Output of ML_1 Model Estimation ... 122

Appendix 4. 20 Stata Output of Wilcoxon Rank-sum (Mann-Whitney) Test of Rubber Quantity Purchased Variable by Transmigrant Status ... 124

(13)

1

Chapter 1

General Introduction

1. General Introduction

(14)

2 Trade dynamics are inseparable from the influence of policies, prices, and the role of traders themselves. Trade policies differ between exporting and importing countries (Reed, 2001), with many trade policies supporting domestic markets while simultaneously threatening the foreign markets with whom they are engaged in trade. The impact of the importing countries’ policies can be detrimental to the domestic trade sector of the exporting countries and vice versa; one common example is import tariffs created by the importing country to protect its domestic industries which subsequently limit the import quantities produced by exporting countries (Reed, 2001). This impact can usually be observed through changes in domestic price trends due to the emergence of a policy, so research aiming to predict the effects of these policies and subsequent price trends on domestic markets can help governments to be better prepared in facing challenges which will inevitably emerge.

Furthermore, changes in price trends can affect the decisions of traders to remain in or exit a market, and ultimately change the structure of the market itself. Due to increased market competition, higher numbers of traders typically result in more favourable the market conditions, while lower numbers of traders typically mean a higher potential for disproportionate market power (Mas-Colell, Whinston, & Green, 1995). In the agricultural sector of developing countries, traders are able to bridge the gap between factories and farmers due to farmers’ lack of capital, information and knowledge (Kopp, Alamsyah, Fatricia, & Brümmer, 2014; Zúñiga-Arias, 2007). Understanding traders’ behavior is relevant for policy makers in suitably anticipating market structure changes and can help protect farmers by maintaining their market power.

Additionally, as we show in chapter 3, another factor affecting traders’ behavior to remain in or exit a market is credit provision. In the agricultural sector, it is a common practice to provide credit for suppliers or farmers, since they still depend on loans, not only for farming activities but also for their daily needs. Carranza and Niles (2019) found that food, agricultural and livestock inputs, and medical expenses are the main loan-dependent expenses among smallholder farmer households. Money lenders certainly carry the risk of losing their money to suppliers who default on their debt, especially when price trends are declining. This may affect traders’ decision in determining a price. Observing the effects of individual loan quantities on price determination then becomes very compelling.

To achieve these objectives, we study the case of oil palm and rubber traders in Jambi province, Indonesia. Indonesia represents a largely agricultural country, with an agricultural sector that accounted for nearly 13% of total Gross Domestic Product in 2018 (Statistics Indonesia, 2019). Among its agricultural outputs, Indonesia’s oil palm and rubber trade

(15)

3 provide a large contribution to the country's foreign exchange (Directorate General of Estate Crops, 2017a, 2017b). Plantation land area in Jambi province, located on Sumatra island, has rapidly expanded, bringing with it many indirect land use change issues (Directorate General of Estate Crops, 2015, 2016a, 2016b, 2017a, 2017b). However, all stakeholders related are benefiting from the sector due to higher income (Bou Dib, Krishna, Alamsyah,

& Qaim, 2018). Moreover, the Indonesian palm oil and rubber industries have resulted in many new domestic employment opportunities. Labor usage within the Palm Oil and Rubber industries increased by about 36 % and 7 %, respectively, from 2013 to 2018 (Directorate General of Estate Crops, 2015, 2016a, 2016b, 2017a, 2017b). The contrasting trajectories of these two valuable industries makes studies related to palm oil and rubber of particular interest to policy makers and national governments.

The research objectives of this study are therefore to:

a. Observe the effects of an importing country’s trade policy on price in a targeted exporting country

b. Analyse factors affecting local traders’ decisions to remain in or exit the market

c. Investigate the influence of farmer’s debt on local traders’ buying choices and price determination

We pursue these research objectives through three separate papers, summarized below:

Paper 1

The European Union’s Biodiesel Antidumping Duty (AD) is one of the most hotly debated international biodiesel trade policies in existence today. In 2013, the EU imposed a biodiesel AD on exporting countries known to engage in biodiesel dumping, and Indonesia was one of the main countries affected The EU accused Indonesian biodiesel producers of charging artificially lower prices than the world market in the purchasing of raw materials (CPO), which was said to affect the performance of EU biodiesel producers. To the best of our knowledge, this is the first study specifically focusing on the price effects of the EU ADs in countries targeted by the duties. The study aims to observe the effects of the AD on Indonesian exports and local CPO prices by applying a Vector Error Correction Model (VECM) approach to time series data. Result shows that the implementation of the AD has a negative effect on the price of Indonesian CPO and oil palm FFB in Jambi province.

(16)

4 Paper 2

Traders play a significant and often underestimated role in agricultural trading activities. Having more traders in the market is favorable for competition, as farmers have more choices as to whom they may sell their products and can choose their most preferred trader. Understanding this role is crucial in understanding the trader’s behavior in the market.

Some may remain, and some may exit the market, which can then in turn alter the market structure. Therefore, this study focuses on traders’ decisions to remain in or exit the market, and the factors influencing this decision. To analyse the probability they remain in the market, we employ a binary logistic regression method to key variables obtained from a 3- round data collection process in Jambi. We find clear evidence that human capital (education and experience), trading structure (traded product, credit provision, land area, operational vehicle ownership, and trader status), structural environment (number of competitors), and socioeconomic (trading revenue) factors all affect the decision of traders to remain in or exit the market.

Paper 3

In the Jambi rubber trade, it is a common practice to reduce the price of rubber to compensate for a contaminated product. However, a farmer’s dependence on loans offered by traders may also have an influence on this price reduction. This study aims to observe to what extent, if any, price reduction, rubber quality, and farmers’ debt influence rubber traders’ preferences in buying rubber, and to estimate how much of a price reduction traders will charge to obtain higher quality rubber or lower farmer’s debt. Using data from 210 rubber traders in Jambi Province, Indonesia, we apply a Discrete Choice Experiment and a willingness-to-pay measurement approach to observe this phenomenon. To the best of our knowledge, there is no study implementing this method to capture agricultural traders’ or middlemen’s behavior in sourcing product and using price reduction as a replacement for willingness to pay; therefore, the methods applied in this study are novel. Result shows that price reduction, rubber quality, and farmer’s debt influence traders’ preferences in buying rubber, whereas higher price reduction, higher rubber quality, and lower farmer’s debt will increase a trader’s utility.

(17)

5

Chapter 2

Does the Biodiesel EU Antidumping Duty Affect the Indonesian CPO Price?

2. Does the Biodiesel EU Antidumping Duty Affect the Indonesian CPO Price?

Rakhma Melati Sujarwo , Bernhard Brümmer , Thomas Kopp

(18)

6 2.1. Introduction

In 2003, the European Union (EU) issued a directive advising member states to increase the use of biofuels and other renewable fuels in the transport sector (European Commission, 2003) in an effort to reduce CO2 emission in transportation. Biodiesel¹ is the most widely used type of biofuel in the EU, with a share of 80.7 % of the market in 2017 (Tiseo, 2019). A study performed by (Sheehan, Camobreco, Duffield, Graboski, & Shapori, 1998) reported that blending higher amounts of biodiesel into petroleum diesel leads to a lower amount of net CO2 life cycle emissions overall² (Figure 2. 1).

Source: Own production based on data from (Sheehan et al., 1998).

Figure 2. 1 Net CO2 Life Cycle Emissions³ of Petroleum Diesel and Biodiesel Blends⁴ The directives aim for new renewable energies to account for at least 32% of total energy needs in the EU by 2030 (European Commission, 2018). The directives have been further expanded a number of times. One of these changes, issued in 2009, requires that renewable fuel be applied not only to motor vehicles, but also to machinery (Bourguignon, 2015). The directives have caused both an increase in the consumption of biofuels and an expansion of the domestic biodiesel industry; however, this growth in domestic consumption has not entirely translated into production increases, and as of 2011 production quantity was still 21.5% lower than consumption (Flach, Lieberz, Lappin, & Bolla, 2018) (Figure 2. 2).

The remaining domestic demand was met through imports.

1 Biodiesel is a renewable diesel fuel substitute made from natural oil or fat mixed chemically with alcohol (Sheehan et al., 1998).

2 Net CO2 was calculated by setting biomass CO2 emissions from the tailpipe to zero (Sheehan et al., 1998).

3 The value is expressed in grams per brake horsepower-hour (g/bhp-h) which is the standards for heavy duty engines (Sheehan et al., 1998).

4 B20 and B100 are biodiesel blends that are respectively 20 and 100 percent biodiesel mixed with petroleum diesel.

(19)

7 Source: Own production based on Flach et al. (2018) and Flach, Lieberz, Rondon, Williams, & Wilson (2016).

Figure 2. 2 Biodiesel and Renewable Diesel (HVO⁵) in the European Union

Opening international trade channels for biodiesel commodities put the European producers under competitive pressure from foreign imports. Indonesia is one of the world’s leading biodiesel exporters, contributing to around 30% of total biodiesel imported by the EU in 2012 (Flach et al., 2018; UFOP, 2017). Biodiesel prices charged by Indonesia are very competitive, since it has a competitive advantage in sourcing Crude Palm Oil (CPO), one of the main raw materials used in producing biodiesel

To confront such a challenge, the EU may enforce certain trade barriers to protect domestic producers. According to Reed (2001), there are four reasons a country or region may have such barriers. First, the government may seek to gain more income by charging tariffs to exporters, which are easier to implement and more discrete than raising income taxes outright. Second, the government may attempt to protect certain products, such as staple foods, to achieve self-sufficiency, and price distortions through trade barriers can be a viable option to this end. Third, the government may attempt to politically protect domestic producers from international rivalry in order to obtain higher subsequent bargaining power in the international market. Lastly, very large importing countries may take advantage of their high market power by restricting imports of certain products to reduce the world price;

thereby increasing their welfare.

One of the most debated international biodiesel trade policies created by the EU is the Biodiesel EU Antidumping Duty (AD) created in late 2013 (European Commission, 2013a). The EU imposed a biodiesel AD on exporting countries known to engage in biodiesel dumping, with Indonesia being one of the main targets of the policy. The country

5 Hydrogenated Vegetable Oil

(20)

8 was accused of having set export prices below the competitive market price, which negatively affected domestic producers’ performance.

This presumption was based on a 2012 study conducted by the EU, which found that Indonesian biodiesel producers were charged far below the world market price in the purchasing of raw materials from Indonesian CPO producers (European Commission, 2012).

The price difference was developed by imposing high-value export tariffs on Indonesian CPO inputs, while maintaining very low export tariffs for the output product of Indonesian biodiesel. The margins generated on dumping were revealed to be between 8.8% and 23.3%

(European Commission, 2013b).

In response, Indonesia made an appeal against the policy at the World Trade Organization’s (WTO) Dispute Settlement Body (DSB) in 2015, as its discussions with the EU had failed to reach a resolution (World Trade Organization, 2018). The court ruled in favor of Indonesia, arguing that the EU failed in their assessment of the situation through inaccurate estimation of production costs, improper formulation of an export price, and insufficient evidence for the existence of a price undercutting scheme. The policy was terminated by the end of 2018, and as of today there are no more ADs imposed by the EU on Indonesian biodiesel (European Union External Action, 2018).

The implementation of trade duties can have important and long-lasting economic effects on stakeholders in target countries. Measuring the effects of the biodiesel AD on the Indonesian CPO industry can provide an important analysis into the economic impacts of trade duties and allow for a quantitative assessment of the effects of ADs on agricultural markets, particularly in developing economies. To the best of our knowledge, there are few studies focusing on the effects of AD’s, especially those imposed by the EU, on target countries (Cheong, 2007; Cuyvers & Dumont, 2005; Jabbour, Tao, Vanino, & Zhang, 2009).

Most studies focus on the effects of AD’s in home countries (Asche, 2001; Avşar, 2013;

Konings & Vandenbussche, 2013; Pierce, 2011), with a primary focus on trade destruction and diversion as a result of duty implementation.

The preceding studies (Cheong, 2007; Cuyvers & Dumont, 2005; Jabbour et al., 2009) applied Ordinary Least Square (OLS) or Propensity Score Matching (PSM) to identify the EU AD’s effects on target countries’ export growth by measuring export value and volume. Other studies (Brambilla, Porto, & Tarozzi, 2012; Chandra & Long, 2013; Lu, Tao,

(21)

9

& Zhang, 2013) concentrated on the effects of United States AD’s on income and labor productivity in target countries by applying Cross-section and Panel Regression Models.

Even though biodiesel usage may lead to decreased CO2 emissions, the expansion of palm oil production land for the biodiesel industry may have detrimental environmental effects, such as biodiversity loss and indirect land use change (Croezen, Bergsma, Otten, &

van Valkengoed, 2010; Matsuda & Takeuchi, 2018). Indirect land use change involves the conversion of land from its initial purpose, such as protected forest, grassland, pasture, or agricultural land, into arable land used for biofuel feedstock cultivation, which results in an increase in emissions (Croezen et al., 2010). This indirect increase in emissions is one of the reasons why biodiesel made from palm oil is currently considered unsustainable by the EU⁶. (European Commission, 2019).

On the positive side, the Indonesian palm oil industry has provided many employment opportunities in the national economy. There was a massive increase in labor usage in the palm oil production sector of about 78% from 2012-2014 (Directorate General of Estate Crops, 2016a). Furthermore, the industry has boosted the Indonesian national income. The value of Indonesian biodiesel exports reached around 1.1 billion USD in 2014, a 17% increase from 2012 (UN Comtrade, 2019).

This contrast between potential environmental benefits and ecological harm make any study of the palm oil industry both interesting and complex. This study differs from previous literatures in its aim to observe the effects of the EU AD on Indonesian exports and local CPO prices by employing time series data to a Vector Error Correction Model (VECM). By comparing Johansen and Gregory-Hansen tests in our model specification, we can prove whether there is a structural break representing the AD duty implementation which captures the effect of the policy implementation on both Indonesian CPO prices and local prices

The study is structured as follows: the subsequent section provides an overview of the alleged dumping occurring in Indonesia and subsequent EU response to the situation.

The model specification section describes the methodology implemented and is followed by a description of the data used in the study. We discuss the results before summarizing our study in the conclusion section.

6 The EU stated that 45% of the oil palm area expansion caused forest devastation; in contrast, that the case for only 8% and less for other biodiesel sources (European Commission, 2019).

(22)

10 2.2. Background: dumping and retaliation

2.2.1 Dumping

The implementation agreement of Article VI of the 1994 General Agreement on Tariffs and Trade expressed an opposition to the practice of dumping (World Trade Organization, 1994). However, many countries still engage in the practice with various products (Figure 2. 3). From an international trade perspective, dumping is associated with a form of price discrimination wherein exporting countries charge a lower price on the world market than they do domestically, or evaluate a product at a value lower than the product’s average cost (Feenstra & Taylor, 2014).

Source: Own production based on data from World Trade Organization (2019).

Figure 2. 3 AD Initiations between 1995-2018.

However, Lindsey and Ikenson (2003) argue that a country can be accused of dumping even when charging equal prices on both the world market and the domestic market; or even a higher world price than domestic price. According to the authors, reasons for this include the effect of fluctuation, the asymmetric treatment of indirect selling expenses, and the exclusion of below-cost sales which are able to influence the value of dumping margins.

To illustrate the general dumping situation, we assume a condition called Foreign Discriminating Monopoly (Feenstra & Taylor, 2014) as shown in Figure 2. 4. The firm is assumed to be in perfect competition on the world market and exercises monopolistic market power on the domestic market. It sets Marginal Cost (MC) equal to Marginal Revenue (MR) to maximize profits. Also, it discriminates between prices set on the local market and those of export markets. To maximize profit, it produces QE, where its local MC meets export

(23)

11 market MRE at point O. However, it sells only at QD to its local market, where MRD meets MREat point N. The difference between QE and QD is the quantity to be exported.

Source: Own production based on data from Feenstra and Taylor (2014).

Figure 2. 4 Dumping Illustration

Additionally, it charges a price PD on the domestic market, while PE is charged on the export market. In this case, PE is lower than PD, even lower than the local Average Cost (AC). This price discrimination is called dumping. In such a condition, nevertheless, the firm still profits. Feenstra and Taylor (2014) mentioned that, even though the export price value is lower than that of average cost (AC), the firm, or the foreign monopoly producer, still benefits from this condition if the export price value is higher than the MC⁷.

However, the alleged dumping scheme exercised by Indonesia is a more complex case. The EU claimed that CPO, the raw material for Indonesian biodiesel, is sold at artificially cheaper prices domestically than internationally (European Commission, 2013b).

This reduces Indonesian biodiesel manufacturers costs below the costs of outside CPO manufacturers, which rely on CPO from Indonesia as an input. The EU further claims that this was caused by a large difference between the value of export tariffs between CPO and biodiesel, where dumping margin arises (European Commission, 2013b). This situation, where there is a difference in the value of export tariffs between input and output products is known as an export tariff escalation (Nogués, 2011).

The objective of imposing an export tariff escalation is to protect the processing industries in the exporting countries (Nogués, 2011). Consequently, the tariff escalation is visible when the raw material export tariff is higher than that of processed product (Corzine,

7 Every unit exported raises profit by the discrepancy between price and MC

(24)

12 2008; Nogués, 2011). Thus, it will encourage the domestic processing industries to grow and increase their competitiveness.

𝑇𝑊 = 𝑇 − 𝑡 (1)

TW = Normal tariff wedge

T = Tariff in ad valorem equivalent of the output commodity t = Tariff in ad valorem equivalent of the input commodity

There is a measurement of tariff escalation, called tariff wedge (TW)⁸ (Elamin &

Khaira, 2003), which is defined as the discrepancy between output and input commodity tariff values (eq. 1). It is more often applied to calculate import than export tariff wedge.

Thus, the interpretation of TW should be reversed, where export tariff escalation appears when TW<0. Meanwhile, export tariff de-escalation occurs when tariff wedge is higher than zero. Another condition called export tariff parity happens when TW=0.

Source: Own calculation based on data from Minister of Trade of the Republic of Indonesia (2019) and Ministry of Finance of the Republic of Indonesia (2019).

Figure 2. 5 Value of Export Tariff (USD/ton) of Indonesian CPO and Biodiesel⁹ Figure 2. 5 depicts the deviation of export tariff values¹⁰ of Indonesian CPO and biodiesel. Since the export tariff value of Indonesian CPO is mostly higher than that of

8 Another measurement, called Effective Rate of Protection, is applied to avoid issue caused by the presence of multiple input and/or output. However, in this case TW is sufficient since CPO is the main raw material in biodiesel production, while other additional materials are negligible.

9 By processing 1 ton of CPO, 0.9 ton of biodiesel is produced (Andarani, Nugraha, & Wieddya, 2017)

10 The CPO export tariff value was calculated by multiplying its export tariff with its Export Standard Price (ESP), where similar process was conducted for biodiesel. The ESP for both CPO and biodiesel were defined based on the CPO Free on Board (FOB) average price. Both CPO and biodiesel export tariffs refer to the CPO reference price which were based on Cost, Insurance, and Freight (CIF) Rotterdam CPO average price and Malaysian and/or Jakarta exchange CPO average price. The export tariff classification regulations were determined by the Ministry of Finance of the Republic of Indonesia. Each is valid until the Ministry decides

(25)

13 Indonesian biodiesel, it can be concluded that the tariff wedge is mostly lower than 0 in the period of 2009-2013. This condition indicates that the export tariff escalation exists. In addition to that, the figure only presents the condition before the EU AD duty imposition, which was started at the end of 2013.

The export tariff imposition will create higher MC for the CPO producers overseas.

Since the domestic CPO price is below the world CPO price due to the export tariff, the MC in exporting biodiesel by Indonesia (MCEX) becomes lower than that incurred by the EU (MCEU) (Figure 2. 6), which is also due to biodiesel low export tariff¹¹. This difference moves the equilibrium along the demand curve (DEU); from point O to B. In response to this new price (PEX), the quantity produced by the EU is reduced from QEU1 to QEU2, where the EU also starts to import QEX-QEU2. This artificially low MCEX means this situation was previously counted as dumping, which was also supported by the dumping margin calculation by the EU. It is consistent with the previous statement by Lindsey and Ikenson, (2003) that even though the world price is higher than the domestic price, a country can still be accused of dumping.

Source: Own illustration.

Figure 2. 6 Indonesian Biodiesel Dumping Illustration 2.2.2 Retaliation

To tackle dumping, the EU imposed an AD. The case was initiated in August 2012.

It was a follow up from the complaint the European Biodiesel Board’s (EBB) filed one month before. The EU biodiesel producer who filed the case represents more than 60% of the total EU biodiesel production (European Commission, 2013a). The evidence provided

that the new regulation is required to be authorized. On the other hand, the CPO reference price and ESP were determined by the Ministry of Trade of the Republic of Indonesia on a monthly basis (Minister of Trade of the Republic of Indonesia, 2019; Ministry of Finance of the Republic of Indonesia, 2019).

11 The low biodiesel export tariff creates the gap between export market MCEX and local MCDOM.

(26)

14 was deemed by the EU to be sufficient to initiate a proper investigation into a potential a dumping scenario. Then, the EU began an investigation into Indonesian exporting producers as well as EU producers, and both parties were required to provide data to be processed as part of the investigation.

Further, dumping and injury margins were examined, and both were compared to determine the AD rates (Table 2. 1), which were imposed at the end of 2013 (European Commission, 2013a). These rates were adjusted to the level of the purity of the biodiesel imported. Additionally, according to Feenstra and Taylor (2014) there are three ways to determine the AD rate: by (1) comparing the import price to the exporter’s local price, (2) comparing this to that of a similar product in another country, and (3) comparing that to the exporter’s AC.

Table 2. 1 The definitive EU AD rate for Indonesian biodiesel producers.

Company Dumping

Margin

Injury

Margin AD rate

PT Ciliandra Perkasa, Jakarta 8.8% 19.7% 8.8%

(EUR 76.94)

PT Musim Mas, Medan 18.3% 16.9% 16.9%

(EUR 151.32) PT Pelita Agung Agrindustri, Medan 16.8% 20.5% 16.8%

(EUR 145.14) PT Wilmar Bioenergi Indonesia, Medan; PT

Wilmar Nabati Indonesia, Medan 23.3% 20.0% 20.0%

(EUR 174.91)

Other cooperating commpanies 20.1% 18.9% 18.9%

(EUR 166.95)

All other companies 23.3% 20.5% 20.5%

(EUR 178.85)

Source: (European Commission, 2013a).

In the same period, the EU imposed the biodiesel AD not only on Indonesia but also on Argentina, where the value of dumping and injury margins, as well as AD rates, are higher for Argentina than they are for Indonesia. The AD rates for Argentina were between 22.0- 25.7% (European Commission, 2013a). Additionally, the United States (US) has also imposed the AD on both Argentina and Indonesia since April 2018, when the International Trade Commission (ITC), an independent US federal trade regulations agency, affirmed that the industry in the US was negatively affected as a result of dumping (Smith, 2018). The

(27)

15 estimated weighted-average dumping margins are much higher than those implemented by the EU.

In the EU, the AD causes an increase in import prices as illustrated in Figure 2. 7, where the price increases from Pw to Pw + duty. When trade is introduced, the Home Consumer Surplus (CS) expands excessively, while Home producers are worse off.

However, there is an increase in the Home total surplus, where area AOB (No Trade) shifts into area ABCD (Free Trade). Under this condition, the import quantity becomes Q3-Q2, while home producers produce only Q2. If producers charge a price higher than Pw, the consumers will import the product.

In a case where the government imposes an AD (Free Trade with AD), the price will increase to Pw + duty. Consequently, the import quantity will fall to Q5-Q4. The Home CS decreases, while there is surplus transfer from consumer to producer. Apart from that, the government will gain revenue from the duty not exceeding the area of EE1FF1. In this case, area CEE1+DFF1 is Dead Weight Loss, and is no longer part of the total surplus. Even though the total surplus decreases by issuing an AD, the government still considers that imposing this policy is beneficial for the reasons mentioned in the previous chapter (Reed, 2001). In this situation, the EU seeks to protect its domestic biodiesel producers from dumping.

Source: Own production based on Reed (2001) and Feenstra and Taylor (2014).

Figure 2. 7 Home Equilibrium without Trade, with Trade and with AD in Trade On the other hand, the possible effect that occurs in Indonesia is illustrated in Figure 2. 8, which is derived from the Foreign Discriminating Monopoly curve. When the export price increases to PE + duty, the domestic price increases. Moreover, the quantity produced will be limited to QE2 where QD2 is sold domestically and QE2-QD2 is exported. Both the

(28)

16 quantity sold domestically and exported may be less than before, and are shown by QD and QE-QD, respectively. However, the effect of the biodiesel AD duty on the Indonesian CPO industry will be discussed further in the results section.

Source: Own illustration.

Figure 2. 8 The AD duty Illustration Effect in Indonesia 2.3. Model Specification

We aim to observe the effects of the EU biodiesel AD on Indonesian export prices and local prices by applying a VECM to time series data. To do so, we test for the existence of a structural break (SB) in the long-run relationship between two price time series. We analyze two relationships - the relationship between world CPO price and Indonesian CPO price, as well as the relationship between Indonesian CPO price and the local price of Fresh Fruit Bunches (FFB) of oil palm in Jambi Province. The existence of a structural break (SB) would indicate changes in long- or short-run relationship between prices caused by the duty imposition if a significant SB occurs around the duty imposition date.

The first step is to ensure that all price variables are stationary on the first difference (I(1)) to avoid the problem of a spurious regression. Thus, we employed Augmented Dicky Fuller (ADF), Phillips-Perron (PP), and Zivot-Andrews (ZA) unit root tests to examine the stationarity. ADF and PP test model are shown by equation 2 and 3 (Asteriou & Hall, 2016).

Dickey-Fuller extended their test procedure by suggesting an augmented version of the test which includes lagged terms of the dependent variable to eliminate autocorrelation (Waheed, Alam, & Ghauri, 2007). the unit root presents itself, i.e. the variable is non-stationary, when the null hypothesis cannot be rejected. Therefore, to obtain the I(1) variable, we expect to reject it on the first difference. Another stationarity test is ZA test (eq. 4) (Zivot & Andrews,

(29)

17 1992), which allows for the univariate existence of a SB. The purpose of the ZA test is to ensure the stationarity robustness and test the initial SB indication in each variable.

∆𝑦_𝑡 = 𝑎_𝑜+ 𝛽𝑦_𝑡−1+ 𝛾𝑡 + ∑ 𝛽_𝑖𝛥𝑦_𝑡−𝑖

𝑝

𝑖=1

+ ℯ_𝑡 (2)

∆𝑦𝑡 = 𝑎𝑜+ 𝛽𝑦𝑡−1+ ℯ𝑡 (3)

∆𝑦_𝑡= 𝑎_𝑜+ 𝛽𝑦_𝑡−1+ 𝛾 𝑡 + 𝜁𝐷𝑈_𝑡 + ∑ 𝛽_𝑖𝛥𝑦_𝑡−𝑖

𝑝

𝑖=1

+ ℯ_𝑡 (4)

Next, we must test the cointegration of the two-pair time series to confirm that they hold a long-run relationship. They are cointegrated if both are integrated in the same order, in this case on I(1), and if there is a linear combination of both series on the level I(0) where in this case the 𝑒𝑡 in equation 5 or 6 is stationary. Initially, we conducted Johansen cointegration test to observe the cointegration. In a later step, we performed a Gregory- Hansen (GH) cointegration test to analyze the cointegration allowing for a SB. These results will be compared to identify whether the SB genuinely exists or not, in the form of VECM.

Before we proceed to the cointegration tests, we need to determine the optimum lag order necessary to reduce bias by using model selection criterions in the form of Akaike information criterion (AIC), Bayesian information criterion (BIC), and Hannan-Quinn information criterion (HQIC) (Mills & Prasad, 1992). We consider all criterions for robustness motivation. The difference between these criteria lie in how the number of estimated parameters and observations are penalized (Mills & Prasad, 1992).

The Johansen cointegration test determines the rank of a time series relationship or the number of cointegrating vectors in a bivariate relationship study with only one possible cointegrating vector (eq. 5). The Johansen test is a maximum likelihood method based on specific correlations (Johansen, 1988). Trace and maximum eigenvalue statistics are also approaches to be observed (Asteriou & Hall, 2016; Johansen, 1988).

𝑦1𝑡 = 𝑎1 + 𝛽𝑦2𝑡+ 𝑒𝑡 (5)

Meanwhile, the GH cointegration test is based on ADF and Phillips (Zα and Zt) test statistic to examine the presence of cointegration allowing SB (Gregory & Hansen, 1996).

There are three different possibilities of SB in the cointegration vector (equation 6); these are (1) a level shift (eq. 6a), (2) a level shift and trend (eq. 6b), and (3) a regime shift (eq.

𝐷𝑈 = 1 if 𝑡 > 𝑏𝑟𝑒𝑎𝑘 𝑑𝑎𝑡𝑒 and 0 if otherwise

(30)

18 6c). We determine the best of these a la Gregory and Hansen (1996). The best model is determined by model selection criteria and test statistics and is presented in the results section. However, in this study, we do not consider cointegration with SB in level and trend (eq. 6b) since there is no indication of trend present in the series and there is a common price volatility. Also, the breakpoint is suggested during this test. Nevertheless, we currently have 2 possibilities left of equation 6 (eq. 6a and 6b), since equation 6b is neglected.

𝑦_1𝑡 = 𝑎₁ + 𝑎₂𝐷𝑈_𝑡 + 𝛽𝑦_2𝑡+ 𝑒_𝑡 (6𝑎)

𝑦_1𝑡 = 𝑎₁ + 𝑎₂𝐷𝑈_𝑡+ 𝑎₃𝑇 + 𝛽𝑦_2𝑡+ 𝑒_𝑡 (6𝑏) 𝑦_1𝑡 = 𝑎₁ + 𝑎₂𝐷𝑈_𝑡+ 𝑎₃(𝐷𝑈_𝑡. 𝑦_2𝑡) + 𝛽𝑦_2𝑡+ 𝑒_𝑡 (6𝑐)

𝐷𝑈 = 1 if 𝑡 > 𝑏𝑟𝑒𝑎𝑘𝑝𝑜𝑖𝑛𝑡 𝑎𝑛𝑑 0 𝑖𝑓 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

When the relationships are properly cointegrated, we proceed to estimate and interpret the long-run relationship by using VECM with log-likelihood function. Equations above (eq. 5, 6a, and 6c) capture the two long-run relationships between the two-pair time series which are distinctly presented below (eq. 7, 8a, 8b and 9, 10a, 10b). All price variables, namely world CPO price (𝑃_𝑊), Indonesian CPO price (𝑃_𝐼𝐷), and Jambi FFB price (𝑃_𝐽𝐵), are in the natural logarithm form. However, we consider extra variables, namely export tariff values in USD/ton (𝐸𝑇) and tax levy (𝑇𝐿) in dummy form, in the relationship between 𝑃_𝐼𝐷

and 𝑃_𝐽𝐵, since those extra variables are part of 𝑃_𝐼𝐷. The coefficient 𝛼₁, 𝛼₂, 𝛼₃,𝛽, 𝜉, and 𝜌 are the parameters to be estimated.

𝑙𝑛𝑃_𝐼𝐷 = 𝛼₁₁+ 𝛽₁. 𝑙𝑛𝑃_𝑊+ 𝑒𝑐𝑡 (7)

𝑙𝑛𝑃_𝐼𝐷 = 𝛼₁₂+ 𝛼₂₁𝑆𝐵 + 𝛽₂. 𝑙𝑛𝑃_𝑊+ 𝑒𝑐𝑡 (8𝑎) 𝑙𝑛𝑃𝐼𝐷 = 𝛼1₃+ 𝛼2₂𝑆𝐵 + 𝛼3₁𝑆𝐵. 𝑙𝑛𝑃𝑊+ 𝛽3. 𝑙𝑛𝑃𝑊+ 𝑒𝑐𝑡 (8𝑐)

and

𝑙𝑛𝑃_𝐽𝐵= 𝛼₁₄+ 𝛽₄. 𝑙𝑛𝑃_𝐼𝐷+ 𝜉₁. 𝐸𝑇 + 𝜌₁. 𝑇𝐿 + 𝑒𝑐𝑡 (9) 𝑙𝑛𝑃𝐽𝐵 = 𝛼1₅+ 𝛼2₃𝑆𝐵 + 𝛽5. 𝑙𝑛𝑃𝐼𝐷+ 𝜉2. 𝐸𝑇 + 𝜌2. 𝑇𝐿 + 𝑒𝑐𝑡 (10𝑎) 𝑙𝑛𝑃_𝐽𝐵= 𝛼₁₆+ 𝛼₂₄𝑆𝐵 + 𝛼₃₂𝑆𝐵. 𝑙𝑛𝑃_𝐼𝐷+ 𝛽₆. 𝑙𝑛𝑃_𝐼𝐷+ 𝜉₃. 𝐸𝑇 + 𝜌₃. 𝑇𝐿 + 𝑒𝑐𝑡 (10𝑐)

In addition to that, VECM can segregate the long-run equilibrium, represented by the error correction term (𝑒𝑐𝑡), from the short-run dynamics. Thus, any shock that occurs in a certain period lets both prices adjust to return to the equilibrium (Patterson, 2000). The short- run equations are presented below. Equation 11a and 11b represent the short-run equilibrium of the relationship between 𝑃_𝐼𝐷and 𝑃_𝑊, while equation 12a and 12b represent the relationship

(31)

19 between 𝑃𝐽𝐵and 𝑃𝐼𝐷. Subscript 𝑡 represents time; 𝑛 describes the number of lags (0, …, 𝑘);

and 𝛾, 𝛿, and 𝜃 are parameters to be estimated.

∆𝑙𝑛𝑃_{𝐼𝐷 𝑡}= ∑ 𝛾₁_𝑛∆𝑙𝑛𝑃_{𝐼𝐷𝑡−𝑛}

𝑘

𝑛=1 + 𝛿₁∆𝑙𝑛𝑃_{𝑊𝑡−𝑛}+ 𝜃₁𝑒𝑐𝑡_𝑡−1 (11𝑎)

∆𝑙𝑛𝑃_{𝑊 𝑡} = ∑ 𝛾₂_𝑛∆𝑙𝑛𝑃_{𝐼𝐷𝑡−𝑛}

𝑘 𝑛=1

+ 𝛿₂∆𝑙𝑛𝑃_{𝑊𝑡−𝑛}+ 𝜃₂𝑒𝑐𝑡_𝑡−1 (11𝑏)

and

∆𝑙𝑛𝑃_{𝐽𝐵 𝑡} = ∑ 𝛾₃_𝑛∆𝑙𝑛𝑃_{𝐽𝐵𝑡−𝑛}

𝑘

𝑛=1 + 𝛿₃∆𝑙𝑛𝑃_{𝐼𝐷𝑡−𝑛}+ 𝜃₃𝑒𝑐𝑡_𝑡−1 (12𝑎)

∆𝑙𝑛𝑃𝐼𝐷 𝑡 = ∑ 𝛾4_𝑛∆𝑙𝑛𝑃_{𝐽𝐵𝑡−𝑛}

𝑘

𝑛=1 + 𝛿4∆𝑙𝑛𝑃_{𝐼𝐷𝑡−𝑛}+ 𝜃4𝑒𝑐𝑡𝑡−1 (12𝑏)

Additionally, the VECM can be briefly presented in matrix notation, as presented below. Equations 19a and 19b represent the two-pair time series relationships without a SB, where the long-run equation is inserted as 𝑒𝑐𝑡 within the short-run equation. Matrices A and B denote the effects of prices in the previous periods. Matrix

ε

denotes normally distributed errors. Equations 20a and 20b define the two-pair time series relationships with SB, where there is a level shift in the model, while equations 21a and 21b define where there is a regime shift in the model. Those equations are differed into parts with a dummy 0 (before SB or t ≤ breakpoint) and dummy 1 (after SB or t > breakpoint).

First, one cointegration vector model allowing for a SB is chosen between two; thus, there is only one model left representing the SB (eq. 20a and 20b or 21a and 21b). Then, that model is compared to the model without the SB (eq. 19a and 19b) to confirm that the effect of SB is legitimate. Again, to do so, we compare the model-selection criteria value for both models: the lower the value, the better the model. Also, the higher the log-likelihood value, the better the model. Another criterion is that when all the GH cointegration test possibilities provide significant results, it is most likely that the model with the SB fits best.

(32)

20

[∆𝑙𝑛𝑃𝐼𝐷

∆𝑙𝑛𝑃_𝑊] = ∑ [𝛾_1𝑛 𝛿₁

𝛾_2𝑛 𝛿₂] [∆𝑙𝑛𝑃_{𝐼𝐷𝑡−𝑛}

∆𝑙𝑛𝑃_{𝑊𝑡−𝑛}]

𝑘

𝑛=1

+ [𝜃1₁

𝜃₂₁] [𝑙𝑛𝑃_𝐼𝐷− 𝛼₁₁− 𝛽₁. 𝑙𝑛𝑃_𝑊] + [𝜀_1𝑡

𝜀_2𝑡] (19𝑎)

[∆𝑙𝑛𝑃𝐽𝐵

∆𝑙𝑛𝑃_𝐼𝐷] = ∑ [𝛾_3𝑛 𝛿₃

𝛾_4𝑛 𝛿₄] [∆𝑙𝑛𝑃_{𝐽𝐵𝑡−𝑛}

∆𝑙𝑛𝑃_{𝐼𝐷𝑡−𝑛}]

𝑘

𝑛=1

+ [𝜃3₁

𝜃4₁] [𝑙𝑛𝑃_𝐽𝐵− 𝛼₁₄− 𝛽₄. 𝑙𝑛𝑃_𝐼𝐷− 𝜉₁. 𝐸𝑇 − 𝜌₁. 𝑇𝐿 ] + [𝜀_3𝑡

𝜀_4𝑡] (19𝑏)

;

[∆𝑙𝑛𝑃_𝐼𝐷

∆𝑙𝑛𝑃_𝑊] =

A + [𝜃1_2𝑎

𝜃₂_2𝑎] [𝑙𝑛𝑃_𝐼𝐷− 𝛼1_2𝑎− 𝛼2_1𝑎𝑆𝐵 − 𝛽2𝑎. 𝑙𝑛𝑃𝑊] + ε , SB = 0 A + [𝜃₁_2𝑏

𝜃2_2𝑏] [𝑙𝑛𝑃_𝐼𝐷− 𝛼₁_2𝑏− 𝛼₂_1𝑏𝑆𝐵 − 𝛽_2𝑏. 𝑙𝑛𝑃_𝑊] + ε , SB = 1

(20𝑎)

[∆𝑙𝑛𝑃_𝐽𝐵

∆𝑙𝑛𝑃_𝐼𝐷] =

B + [𝜃3_2𝑎

𝜃₄_2𝑎] [𝑙𝑛𝑃_𝐽𝐵− 𝛼1_5𝑎− 𝛼2_3𝑎𝑆𝐵 − 𝛽5𝑎. 𝑙𝑛𝑃𝐼𝐷− 𝜉2𝑎. 𝐸𝑇 − 𝜌2𝑎. 𝑇𝐿 ] + ε , SB = 0 B + [𝜃₃_2𝑏

𝜃4_2𝑏] [𝑙𝑛𝑃_𝐽𝐵− 𝛼₁_5𝑏− 𝛼₂_3𝑏𝑆𝐵 − 𝛽_5𝑏. 𝑙𝑛𝑃_𝐼𝐷− 𝜉_2𝑏. 𝐸𝑇 − 𝜌_2𝑏. 𝑇𝐿 ] + ε , SB = 1

(20𝑏)

;

[∆𝑙𝑛𝑃_𝐼𝐷

∆𝑙𝑛𝑃𝑊] =

A + [𝜃1_3𝑎

𝜃₂_3𝑎] [𝑙𝑛𝑃𝐼𝐷− 𝛼1_3𝑎− 𝛼2_2𝑎𝑆𝐵 − 𝛼3_1𝑎𝑆𝐵. 𝑙𝑛𝑃𝑊− 𝛽3𝑎. 𝑙𝑛𝑃𝑊] + ε , SB = 0 A + [𝜃1_3𝑏

𝜃₂_3𝑏] [𝑙𝑛𝑃_𝐼𝐷− 𝛼₁_3𝑏− 𝛼₂_2𝑏𝑆𝐵 − 𝛼₃_1𝑏𝑆𝐵. 𝑙𝑛𝑃_𝑊− 𝛽_3𝑏. 𝑙𝑛𝑃_𝑊] + ε^, ^{SB = 1}

(21𝑎)

[∆𝑙𝑛𝑃𝐽𝐵

∆𝑙𝑛𝑃𝐼𝐷] =

B + [𝜃₃_3𝑎

𝜃₄_3𝑎] [𝑙𝑛𝑃_𝐽𝐵− 𝛼₁_6𝑎− 𝛼₂_4𝑎𝑆𝐵 − 𝛼₃_2𝑎𝑆𝐵. 𝑙𝑛𝑃_𝐼𝐷− 𝛽_6𝑎. 𝑙𝑛𝑃_𝐼𝐷− 𝜉_3𝑎. 𝐸𝑇 − 𝜌_3𝑎. 𝑇𝐿 ] + ε , SB = 0 B + [𝜃3_3𝑏

𝜃₄_3𝑏] [𝑙𝑛𝑃_𝐽𝐵− 𝛼1_6𝑏− 𝛼2_4𝑎𝑆𝐵 − 𝛼3_2𝑏𝑆𝐵. 𝑙𝑛𝑃𝐼𝐷− 𝛽6𝑏. 𝑙𝑛𝑃𝐼𝐷− 𝜉3𝑏. 𝐸𝑇 − 𝜌3𝑏. 𝑇𝐿 ] + ε , SB = 1 (21𝑏)

At last, if the SB effect is legitimate, we compare the full model (also called the restricted model) which allows for the presence of the SB (eq. 20 or eq. 21) to the unrestricted model (also called as separated model). The unrestricted model represents the separation of equation 19 into before and after the breakpoint, thus we will have two long-run relationships (two-unrestricted model). To do so, a Likelihood Ratio (LR) test is required, since the VECM model determination employs the log-likelihood function. The log-likelihood discrepancy between a full model and the two-unrestricted model determines the LR value (eq. 22), which is then compared with a chi-square (χ²) distribution (Wooldridge, 2013). The two- unrestricted model is better than the full model if we can reject the null hypothesis.

𝐿𝑅 = 2 (𝐿_𝑢𝑟− 𝐿_𝑟) (22)

𝑒𝑐𝑡

A

B

𝛆

Palm Oil and Rubber Price and Trader’s Behavior at International towards Local Level